Detectors designed for infrared imaging are conventionally produced as a one or two-dimensional array of elementary detectors, or bolometers, said bolometers taking the form of membranes suspended above a substrate which is generally made of silicon, by means of support arms that have a high thermal resistance.
The substrate usually incorporates means of sequentially addressing the elementary detectors and means of electrically exciting and pre-processing the electrical signals generated by these bolometers. This substrate and the integrated means are commonly referred to as the “readout circuit”.
In order to obtain a scene infrared image using this detector, the scene is projected through suitable optics onto the array of bolometers and clocked electrical stimuli are applied via the readout circuit to each of the bolometers or to each row of such bolometers in order to obtain an electrical signal that constitutes an image of the temperature reached by each of said elementary detectors. This signal is then processed to a greater or lesser extent by the readout circuit and then, if applicable, by an electronic device outside the package in order to generate a thermal image of the observed scene.
This type of detector has numerous advantages in terms of its manufacturing cost and implementation but also has drawbacks that limit the performance of systems that use such detectors. In particular, there are problems with regard to the uniformity of the image obtained. In fact, when exposed to a uniform scene, not all the bolometers respond in exactly the same way and this results in fixed spatial noise in the image thus obtained.
This variability has several sources. In particular, technological variability of the resistance of the bolometers causes, among other defects, offset variation and gain variation in the image, i.e. in the case of offset, spatial variation in the output levels of bolometers exposed to a uniform image and, in the case of gain, variability of the absolute variation in the output levels of bolometers that are exposed to a uniform temperature variation of a scene.
There are numerous offset correction methods. A first method for correcting offset variation involves using offset correction tables which are prepared after factory calibration operations. However, the stability of these corrections depends on the temperature stability of the focal plane and thus, in non-temperature controlled applications (commonly referred to as “TEC-less”), it is necessary to resort to acquiring and storing gain and offset tables for multiple, so-called calibration temperatures and then using said tables when the detector is operated, for instance by interpolation, in order to ensure continuous digital correction over the entire operational dynamic range, in terms of temperature, of the focal plane of the detector. These tables, obtained using a factory calibration test bench, incur significant costs for manufacturers, especially in terms of the equipment that is installed in the calibration test bench and the time it takes to acquire all the gain and offset tables.
Another method, disclosed for example in Document US 2002/0022938, involves acquiring an image of a uniform reference scene by closing a mechanical shutter. Once this image has been acquired, the shutter is opened and the reference image is stored and then digitally or analogically subtracted from the current images. This method is more widely known as “shutter correction” or “one-point correction”. It has the advantage of enabling highly efficient correction at around the ambient temperature of the detector which was used to acquire the reference image and requires little memory and few computing resources.
On the other hand, this method involves using a mechanical shutter—a mechanical device which has a non-negligible cost, is relatively fragile because of the moving parts it contains and consumes energy. What is more, if operating conditions change and, more especially the thermal environment of the detector changes, the images acquired from the scene deteriorate due to the reappearance of offset variation and it is then necessary to acquire a reference image again by closing the mechanical shutter. In fact, the detector is unusable, at least for the time it takes to acquire the reference image.
Another offset correction method which is disclosed, for example, in document WO 98/47102, involves digitally processing a series of consecutive images contained in a rolling time window that includes enough frames to make it possible to extract a continuous component from the time window. The spatial distribution of this continuous component, which is similar to the offset distribution, is then digitally subtracted from the current acquired images.
However, this not only suppresses the actual offset variation itself, it also suppresses all the static information from the scene. There is admittedly no need to use a mechanical shutter but offset correction like this is only really acceptable in cases where the scene is essentially permanently variable or moving. In fact, all details and fixed local contrasts over a duration that is equal to or greater than the duration of the rolling window are processed as non-uniformities and, by virtue of this, are corrected in the same way as fixed spatial noise.
Generally speaking, offset correction methods according to the prior art are only applied once an image has been acquired and therefore correct the effects of offset variation in the image. Nevertheless, although offset variation impacts image quality as such due to the presence of noise that is independent of the scene, it also has effects on the dynamic range of the observable scene that these types of techniques do not correct.
To correct this phenomenon, FIG. 1 shows a basic detection and readout layout of the kind that is conventionally used in bolometric array detectors.
This basic layout comprises:                a picture element, or pixel, 10, comprising, in particular, an imaging bolometer 12 and components 14 and 16 that are needed in order to implement it;        an integrating circuit 18 used to read imaging bolometer 12; and        a compensation circuit 20 for compensating a common-mode current that flows through imaging bolometer 12 when the latter is read.        
Bolometer 12 is subjected to infrared radiation IR originating from a scene and is connected to ground by a first terminal A.
Integrating circuit 18 comprises:                an operational amplifier 22, the non-inverting input (+) of which is kept at a predetermined constant voltage VBUS;        a capacitor 24, having a predetermined capacitance Cint and connected between the inverting input (−) of amplifier 22 and the output of the latter; and        a reset switch 26 connected in parallel with capacitor 24 and controllable by means of a “Reset” signal.        
Picture element 10 also comprises a read switch 16 that can be controlled by means of a “Select” signal and is connected to the inverting input (−) of the operational amplifier and a first MOS injection transistor 14, the gate of which is controlled by a voltage VFID so as to impose a voltage Vac across the terminals of bolometer 12, the source of which is connected to a second terminal B of bolometer 12 and the drain of which is connected to the other terminal of read switch 16.
Compensation circuit 20 used to compensate the common-mode current that flows through imaging bolometer 12 comprises a resistive compensation bolometer 28 made of the same material as imaging bolometer 12. Compensation bolometer 28 is essentially insensitive to radiation originating from the scene, for instance because it has a low thermal resistance relative to the substrate and is, optionally or alternatively, provided with an opaque shield 30.
One of the terminals of compensation bolometer 28 is connected to a predetermined voltage VSK and its other terminal is connected to the source of a second MOS injection transistor 32 of circuit 20. The drain of injection transistor 32 is connected to the inverting input (−) of operational amplifier 22 and its gate is connected to a predetermined voltage GSK.
In order to read bolometer 12, once capacitor 24 has discharged due to zero reset switch 26 closing, imaging and compensation bolometers 12, 28 are biased by the control voltage of biasing transistors 14, 32 and the difference between current lac that flows through imaging bolometers 12 and current Iav that flows through compensation bolometer 28 is integrated by integrating circuit 18 over a predetermined integration duration Tint. As is known in itself, the use of compensation circuit 20 is justified by the fact that the useful current, i.e. that which is representative of the temperature of the scene, only accounts for a minute portion, generally around 1%, of the total current that flows through imaging bolometer 12, hence the need to eliminate the common-mode current before integration.
The voltage Vout on the output of integrator 18 is then given by the equation:
                    Vout        =                  VBUS          +                                    1                              C                int                                      ⁢                                          ∫                0                                  T                  ⁢                                                                          ⁢                  int                                            ⁢                                                (                                                            Iac                      ⁡                                              (                        t                        )                                                              -                                          Iav                      ⁡                                              (                        t                        )                                                                              )                                ⁢                                                                  ⁢                                  ⅆ                  t                                                                                        (        1        )            
Integration by circuit 18 thus makes it possible to apply, through the value of capacitance Cint, gain to readout of the wanted signal whilst ensuring conversion of the useful current to a voltage that is simpler to manipulate. This way, all the imaging bolometers of the array detector are read in the same way, especially by applying the same bias level.
The layout and operation of the components described above is conventional and is not explained in any greater detail for the sake of brevity. For additional details, the reader is advised to consult, for example, the document entitled “Uncooled amorphous silicon enhancement for 25 μm pixel pitch achievement” by E. Mottin et al, Infrared Technology and Application XXVIII, SPIE, vol. 4820.
Assuming, for instance, that the relative spatial variation in the resistance of the imaging bolometers of the detector equals 1%, resulting in, for low bias levels, a 1% variation in currents Iac, and that biasing of the compensation bolometers is selected so that current Iav equals approximately 90% of current Iac, the spatial variation in voltages Vout after all the bolometers have been read is approximately 10%. In conventional detectors, this variation represents around 300 mV of their dynamic output response. If biasing of the imaging bolometers is also increased, for instance by 50%, in order to increase the value of the output levels and hence the sensitivity of the detector, the variation in output voltages Vout also increases by 50% and then reaches 450 mV. Considering that the total dynamic response available is usually limited to 2 or 3 V, a significant portion of this dynamic response is therefore used up by the natural variability of bolometers alone.
Thus, offset variation, simply by existing, uses up a portion of the dynamic output response of a detector. The term “residual dynamic response” or “dynamic scene response” is usually used to denote the difference between the maximum amplitude of voltage Vout when the integrating circuits are not saturated and the maximum amplitude of output voltages Vout when exposed to a uniform scene, i.e. the remaining dynamic response to the wanted signal.
Besides the residual dynamic response being less than the electrical dynamic response of the integrating circuits simply due to the presence of offset variation, this residual dynamic response diminishes as the sensitivity desired by the user increases.
Also, when the imaging bolometers are biased, their temperature rises due to the Joule effect, resulting in increased amplification of variation in the currents that flow through them and hence the output voltages, thus resulting in a reduced residual dynamic response. A similar phenomenon also occurs when the temperature of the focal plane on which the bolometer array is positioned is increased. Because usual bolometric materials have a negative coefficient of thermal resistance, this results in variability of the output levels of the bolometers increasing rapidly, thereby significantly diminishing the residual dynamic response.
It should be noted that offset variation corrections according to the prior art do not deal with this reduction in residual dynamic response in any way and confine themselves to retrospectively correcting the effects of said variability on images that have already been formed.